kaggle 데이터를 이용한 빅마트의 상품 가격 예측

library(kableExtra) 

colSums(is.na(big)) %>%
    kbl(col.names = "결측값 수") %>% 
    kable_styling(bootstrap_options = "striped", full_width = F) #열별 결측값

library(naniar)

gg_miss_var(big %>% select(Item_Weight, Outlet_Size), show_pct = TRUE) +
  scale_y_continuous(limits = c(0,100))+
  labs(x = "", y = "결측값 비율(전체)") +
  theme(axis.text = element_text(size = 12),
        axis.title = element_text(face = "bold", 
                                  hjust = 0.5, size = 15, color = "darkblue"))

big %>%
  select(Outlet_Identifier:Item_Outlet_Sales) %>% 
  ggplot() +
  geom_col(aes(Outlet_Identifier, Outlet_Size, fill = Outlet_Size)) +
  scale_y_discrete(labels = rep("",4))

big %>%
  select(Outlet_Identifier:Item_Outlet_Sales) %>%  
  group_by(Outlet_Type, Outlet_Size) %>% 
  count() %>% 
  kbl() %>% 
  kable_styling(bootstrap_options = "striped", full_width = F) %>% 
  column_spec(2, color = spec_color(c(1,2,1,1,1,2,1,1)))

#마트 규모와 마트 분류의 상관계수
tab <- ftable(big$Outlet_Size, big$Outlet_Type)
library(DescTools) #Association measures {DescTools}
CramerV(tab)

#마트 규모와 마트가 위치한 도시규모의 상관계수
tab2 <- ftable(big$Outlet_Size, big$Outlet_Location_Type)
CramerV(tab2)

big[which(big$Outlet_Type=="Grocery Store" & is.na(big$Outlet_Size)), "Outlet_Size"] <- "Small"

big %>% 
  filter(Outlet_Type == "Supermarket Type1", Outlet_Location_Type == "Tier 2") %>% 
  group_by(Outlet_Type, Outlet_Location_Type, Outlet_Size) %>% 
  summarise(n()) %>% 
  kbl(col.names = c("Outlet_Type", "Outlet_Location_Type", "Outlet_Size","n")) %>% 
  kable_styling(bootstrap_options = "striped", full_width = F)

big[which(big$Outlet_Type=="Supermarket Type1" & is.na(big$Outlet_Size)), "Outlet_Size"] <- "Small"

#colSums(is.na(big))

# 무게가 NA인 상품
NA_Item_Identifier <- big %>% 
  filter(is.na(Item_Weight)) %>% 
  select(Item_Identifier) %>% 
  pull()

# 동일 제품의 평균을 구함
weight_mean <- big %>% 
  filter(!is.na(Item_Weight)) %>% 
  filter(Item_Identifier %in% NA_Item_Identifier) %>% 
  group_by(Item_Identifier) %>% 
  summarise(mean = mean(Item_Weight)) 

#상품 ID 에 대한 평균무게로 결측값 대치
big[which(is.na(big$Item_Weight)), "Item_Weight"] <- big %>% 
  filter(is.na(Item_Weight)) %>% 
  select(Item_Identifier, Item_Weight) %>% 
  left_join(weight_mean, by = "Item_Identifier") %>% 
  select(Item_Weight = mean)

colSums(is.na(big))

# Item_Weight 분포
unimputed_wt <- ggplot(un_big, aes(x=Item_Weight)) +  # 결측값 대치 전
  geom_density(fill="blue", alpha=.5) +
  geom_line(stat="density")

imputed_wt <- ggplot(big, aes(x=Item_Weight)) +   # 결측값 대치 후
  geom_density(fill="blue", alpha=.5) +
  geom_line(stat="density")

library(cowplot)
ggdraw() +             # ggarrange(unimputed_wt, imputed_wt){ggpubr}
  draw_plot(unimputed_wt, x = 0, y = 0, width = .5, height = .95) +
  draw_plot(imputed_wt, x = 0.5, y = 0, width = .5, height = .95) +
  draw_plot_label(label = c("unimputed", "imputed"), size = 15,
                  x = c(0, 0.5), y = c(1, 1))

# Outlet_Size에 따른 가격 분포
library(ggridges)

unimputed_size <- ggplot(un_big, aes(x = Item_Outlet_Sales, y = Outlet_Size, fill = Outlet_Size)) +  # 결측값 대치 전
  geom_density_ridges2(scale=1) +
  theme(legend.position = "none")

imputed_size <- ggplot(big, aes(x = Item_Outlet_Sales, y = Outlet_Size, fill = Outlet_Size)) +  # 결측값 대치 후
  geom_density_ridges2(scale=1) +
  theme(legend.position = "none")

ggdraw() +             # ggarrange(unimputed_size, imputed_size)
  draw_plot(unimputed_size, x = 0, y = 0, width = .5, height = .95) +
  draw_plot(imputed_size, x = 0.5, y = 0, width = .5, height = .95, ) +
  draw_plot_label(label = c("unimputed", "imputed"), size = 15,
                  x = c(0, 0.5), y = c(1, 1))